bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Identification of Fibronectin 1 as a candidate genetic modifier in a Col4a1 mutant mouse 2 model of Gould syndrome 3 4 Mao Mao1, Tanav Popli1§, Marion Jeanne1♯, Kendall Hoff1+, Saunak Sen2,3,4, and Douglas B. 5 Gould1,3,5* 6 7 8 1Departments of Ophthalmology, University of California, San Francisco, San Francisco, CA. 94143, 9 USA; 2Department of Epidemiology and Biostatistics, University of California, San Francisco, San 10 Francisco, CA. 94143, USA; 3Institute of Human Genetics, University of California, San Francisco, 11 San Francisco, CA. 94143, USA; 4Department of Preventive Medicine, University of Tennessee 12 Health Science Center, 66 North Pauline St, Memphis, TN 38163 USA; 5Department of Anatomy, 13 University of California, San Francisco, San Francisco, CA 94143, USA 14 15 *Corresponding author: 16 Dr. Douglas B. Gould, Departments of Ophthalmology and Anatomy, Institute for Human Genetics, 17 UCSF School of Medicine, 10 Koret Way, Room 235, San Francisco, CA, 94143-0730, (415) 476- 18 3592 (telephone), [email protected], ORCID identifier number 0000-0001-5127-5328 19 §Current address: University of Michigan, Department of Neurology, 1500 E Medical Center Drive, 20 Taubman Center, Clinic 1C, Ann Arbor, MI, 48109 21 ♯Current address: Genentech Inc, 1 DNA Way, South San Francisco, CA 94080 22 +Current address: Centrillion Technologies, 2500 Faber Pl. Palo Alto, CA 94303 23 24 25 Key words: 26 Gould syndrome, COL4A1, basement membrane, modifier, fibronectin, integrin signaling 27

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

28 SUMMARY STATEMENT 29 A genetic screen in mice implicates FN1 as a genetic modifier of some, but not all, aspects of Gould 30 syndrome.

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

31 ABSTRACT 32 Collagen type IV alpha 1 and alpha 2 (COL4A1 and COL4A2) are major components of almost all 33 basement membranes. COL4A1 and COL4A2 mutations cause a multisystem disorder called Gould 34 syndrome which can affect any organ but typically involves the cerebral vasculature, eyes, kidneys and 35 skeletal muscles. The manifestations of Gould syndrome are highly variable and animal studies suggest 36 that allelic heterogeneity and genetic context contribute to the clinical variability. We previously 37 characterized a mouse model of Gould syndrome caused by a Col4a1 mutation in which the severities 38 of ocular anterior segment dysgenesis (ASD), myopathy, and intracerebral hemorrhage (ICH) were 39 dependent on genetic background. Here, we performed a genetic modifier screen to provide insight 40 into the mechanisms contributing to Gould syndrome pathogenesis and identified a single 41 (modifier of Gould syndrome 1; MoGS1) on 1 that suppressed ASD. A separate screen 42 showed that the same locus ameliorated myopathy. Interestingly, MoGS1 had no effect on ICH, 43 suggesting that this phenotype may be mechanistically distinct. We refined the MoGS1 locus to a 4.3 44 Mb interval containing 18 coding , including Fn1 which encodes the extracellular matrix 45 component fibronectin 1. Molecular analysis showed that the MoGS1 locus increased Fn1 expression 46 raising the possibility that suppression is achieved through a compensatory extracellular mechanism. 47 Furthermore, we show evidence of increased integrin linked kinase levels and focal adhesion kinase 48 phosphorylation in Col4a1 mutant mice that is partially restored by the MoGS1 locus implicating the 49 involvement of integrin signaling. Taken together, our results suggest that tissue-specific mechanistic 50 heterogeneity contributes to the variable expressivity of Gould syndrome and that perturbations in 51 integrin signaling may play a role in ocular and muscular manifestations.

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

52 INTRODUCTION 53 Collagens are the most abundant in the body, making up ~30% of the dry weight. The collagen 54 superfamily of extracellular matrix (ECM) molecules comprises 28 members encoded by 46 genes 55 (Ricard-Blum, 2011). Among these, type IV collagens are primordial ECM molecules and are 56 fundamental constituents of specialized structures called basement membranes (Fidler et al., 2017). In 57 mammals, six genes encode the type IV collagens (COL4A1 to COL4A6) and their protein products 58 assemble into three distinct heterotrimers (α1α1α2, α3α4α5 or α5α5α6). The α1α1α2 network is 59 ubiquitous throughout development and in most adult tissues and its absence results in embryonic 60 lethality (Poschl et al., 2004). Pathogenic mammalian Col4a1 mutations were first identified using 61 forward mutagenesis screens in mice with variable forms of ocular pathology (Favor et al., 2007; Gould 62 et al., 2005; Thaung et al., 2002) and human mutations were identified in individuals with severe 63 inherited or de novo porencephaly and early onset intracerebral hemorrhages (ICH) (Breedveld et al., 64 2006; Gould et al., 2005; Gould et al., 2006; Sibon et al., 2007; Vahedi et al., 2007). A number of studies 65 in humans and mice have subsequently expanded the phenotypic spectrum and it is now well 66 established that mutations in COL4A1 and COL4A2 cause a multi-system disorder (Jeanne and Gould, 67 2017; Labelle-Dumais et al., 2019; Mao et al., 2015; Meuwissen et al., 2015; Yoneda et al., 2013; 68 Zagaglia et al., 2018) that has adopted the name Gould syndrome 69 (https://www.gouldsyndromefoundation.org). 70 71 Gould syndrome is highly clinically heterogeneous and can have variable penetrance and severity 72 across many organs. Cerebrovascular, ocular, renal and neuromuscular pathologies are among the 73 most commonly described manifestations. We originally described a mouse model of Gould syndrome 74 with a Col4a1 splice acceptor mutation that leads to exclusion of exon 41 (Δex41) from the mature 75 transcript and 17 amino acids from the triple-helical domain of the protein (Gould et al., 2005). However, 76 the majority of disease-causing variants reported in humans are missense mutations in highly 77 conserved glycine residues of the triple-helical domain (Jeanne and Gould, 2017). In order to faithfully 78 replicate human disease, we compared the effects of different mutations in an allelic series composed 79 of nine distinct Col4a1 and Col4a2 mutant mouse strains (Favor et al., 2007; Jeanne et al., 2015; Kuo 80 et al., 2014). We established that allelic heterogeneity has important implications for penetrance and 81 severity of various pathologies including ICH and myopathy. We found that mutations can differ in 82 degree or kind and that allelic differences contribute to clinical variability, in part, by tissue-specific 83 mechanistic heterogeneity (Labelle-Dumais et al., 2019). However, variable clinical manifestations 84 including age of onset and severity among organs have been reported in people with the same recurrent 85 or inherited mutations, and reduced penetrance and asymptomatic carriers exist in some families, 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

86 indicating the involvement of factors in addition to allelic heterogeneity (Coupry et al., 2010; de Vries et 87 al., 2009; Rødahl et al., 2013; Shah et al., 2012). 88 89 Studies using inbred strains of mice clearly demonstrate that a mutation can have variable outcomes 90 in different genetic contexts underscoring the important role that genetic modification plays in clinical 91 heterogeneity. When compared to other mutations in the allelic series, Col4a1Δex41 tends to have the 92 most severe phenotypes across multiple organs when maintained on a pure C57BL/6J (B6) genetic 93 background (Gould et al., 2007; Jeanne et al., 2015; Kuo et al., 2014). However, when the Col4a1+/Δex41 94 B6 mice were crossed for a single generation to the CAST/EiJ (CAST) genetic background, ocular 95 anterior segment dysgenesis (ASD), ICH and skeletal myopathy were all significantly reduced in the 96 F1 progeny (CASTB6F1) (Gould et al., 2007; Jeanne et al., 2015; Labelle-Dumais et al., 2011). 97 Interestingly, when Col4a1+/Δex41 B6 mice were crossed for a single generation to the 129SvEvTac (129) 98 genetic background, ASD was ameliorated (albeit to a lesser extent than in CASTB6F1 mice) but ICH 99 was not (Gould et al., 2007; Jeanne et al., 2015). Collectively, these data suggest that the B6 genetic 100 background confers susceptibility to develop severe Gould syndrome related pathologies and that 101 CAST and 129 strains have one or more locus/loci that can genetically suppress pathology in a tissue- 102 specific manner. 103 104 Identification of genetic modifier genes can help reveal disease mechanisms and potential therapeutic 105 targets. Phenotype-driven mapping studies using mice and other model organisms are a powerful 106 approach to identify genetic interactions that may be difficult to discover in humans even when large 107 pedigrees are available (Ceco and McNally, 2013; Meyer and Anderson, 2017; Vieira et al., 2015). We 108 previously performed a genetic modifier screen for pathology caused by the Col4a1Δex41 mutation to 109 provide insight into the underlying pathogenic mechanisms. Because 1) ASD is relatively severe, easily 110 screened, and robustly suppressed by two strains and 2) the magnitude and breadth of phenotypic 111 rescue was greater in CAST compared to 129 mice, we selected ASD and CAST as the phenotype and 112 strain, respectively, to perform a pilot suppressor screen in a small number of mice (Gould et al., 2007). 113 We generated CASTB6F1 mice and iteratively crossed mutant mice back to the B6 background. In 114 each backcross, we applied selective pressure to retain one or more genetic suppresser locus by 115 choosing the progeny with the mildest phenotype to breed for the next generation. Using a crude 116 genome-wide scan, we identified a dominant modifier locus in CAST Chromosome (Chr) 1 that 117 suppressed ASD. However, the approach had significant limitations including screening on a single 118 phenotype, the inability to identify potential recessive loci and the possibility that we may have 119 incidentally overlooked other dominant loci by imposing bottlenecks at each generation. 5 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

120 121 Here, we performed independent genetic modifier screens for ASD and skeletal myopathy on a large 122 scale F2 cross with the potential to identify multiple dominant or recessive loci that enhanced or 123 suppressed pathology. Both screens identified a single suppressor locus on CAST Chr 1 that we called 124 modifier of Gould syndrome 1 (MoGS1). Surprisingly, although the CAST background also suppressed 125 ICH, the effect was not attributable to the MoGS1 locus. These data suggest that ASD and skeletal 126 myopathy may be mechanistically related to each other but distinct from ICH which further supports the 127 notion of tissue-specific mechanistic heterogeneity contributing to the clinical variability of Gould 128 syndrome (Labelle-Dumais et al., 2019). Furthermore, molecular analyses suggest that the ECM 129 protein fibronectin 1 (FN1) is a strong candidate as the genetic suppressor at the MoGS1 locus and 130 provide evidence for a role of altered integrin signaling in Gould syndrome pathogenesis. 131 132 RESULTS 133 High-resolution genome-wide mapping identified a suppressor locus for Col4a1-related ASD 134 and myopathy on mouse Chr 1. To identify modifiers of Gould syndrome, we first performed a 135 genome-wide screen with the early onset and easily observed ASD phenotype. We phenotyped and +/Δex41 136 genotyped 192 mutant F2 progeny from (CAST X B6-Col4a1 ) F1 intercrosses using a mouse 137 medium density linkage panel with 646 informative SNPs (Fig. 1A-C). We identified a region of interest 138 on CAST Chr 1 with a LOD score of 11.2 and Bayesian confidence interval extending from 51.3 to 73.0 139 Mb (Fig.1C). To determine if smaller effects of other loci might have been masked, we performed a 140 second genome-wide scan conditioned on this locus, but no additional loci reached statistical 141 significance (Fig. 1D). This observation suggests that the major modifying effect of the CAST 142 background on ASD is imparted through this single dominant locus on Chr 1. 143 144 In addition to suppressing ASD, the CAST genetic background significantly suppresses skeletal 145 myopathy in Col4a1 mutant mice (Labelle-Dumais et al., 2011). To identify potential modifier loci for 146 this phenotype, we performed an independent genome-wide mapping analysis using skeletal myopathy 147 as the primary phenotype. We quantified myopathy severity as the percentage of muscle fibers 148 containing non-peripheral nuclei (NPN) in histological sections of quadriceps (Fig. 1E). This assay is 149 more quantitative with greater statistical power but also more labor-intensive and therefore we analyzed +/Δex41 150 a randomly generated subset (n= 49) of the mutant (CAST X B6-Col4a1 ) F2 progeny. The 151 genome-wide scan revealed an interval on Chr 1 defined by the same markers as the ASD modifier 152 locus (max LOD score = 5.61) (Fig. 1F). 153 6 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

154 To validate the biological effect of the MoGS1 locus, we independently generated an incipient congenic 155 strain by iterative backcrossing the CAST-derived locus onto the susceptible B6 genetic background 156 for 5 generations (N5). At N5, the genetic background of the mice is ~97% pure B6 on average, except 157 that they carry the CAST Chr 1 interval that includes the MoGS1 locus. Incipient congenic mice were 158 intercrossed to generate Col4a1+/Δex41 mice with zero (B/B), one (C/B) or two (C/C) copies of the CAST- 159 derived chromosomal interval (Fig. 2). Out of 24 eyes from Col4a1+/Δex41 mice that were homozygous 160 for the B6 allele (MoGS1B/B) at the congenic interval, 29% (7) were moderate, and 71% (17) were 161 severe (Fig. 2A). In contrast, out of 28 eyes from the heterozygous group (MoGS1C/B) there were 21.5% 162 (6), 46.5% (13) and 32% (9) of eyes that were scored with mild, moderate and severe ASD, respectively. 163 In mice homozygous for the CAST allele (MoGS1C/C), there were 50% (10), 35% (7) and 15% (3) of 164 eyes that were mild, moderate and severe, respectively. In a parallel experiment, we tested the effect 165 of this congenic locus on myopathy and observed a dosage effect, whereby Col4a1+/Δex41 mice that 166 were heterozygous at the MoGS1 locus (MoGS1C/B) showed a trend toward reduced myopathy while 167 homozygous MoGS1C/C incipient congenic mice had significantly milder myopathy compared to mice 168 that were homozygous for the B6 allele (MoGS1B/B). Together, these results support the existence of 169 one or more semi-dominant modifier genes at the MoGS1 locus on Chr 1. 170 171 MoGS1 does not reduce porencephaly penetrance or ICH severity in Col4a1 mutant mice 172 Individuals with Gould syndrome have highly penetrant and clinically variable cerebrovascular diseases 173 that include porencephaly and ICH (Bilguvar et al., 2009; de Vries et al., 2009; Giorgio et al., 2015; 174 Vilain et al., 2002). ICH was previously reported to be significantly reduced in Col4a1 mutant mice on 175 a CASTB6F1 background compared to those maintained on a B6 background suggesting that the 176 CAST genetic background also suppresses this phenotype (Jeanne et al., 2015). To test whether the 177 MoGS1 locus might also genetically modify cerebrovascular diseases, we assessed porencephaly 178 penetrance and ICH severity of the incipient congenic mice. We found that porencephaly penetrance 179 and ICH severity were similar among Col4a1+/Δex41 mice irrespective of their genotypes at the MoGS1 180 locus (Fig. 2C, D). 181 182 Fine mapping with subcongenic lines refined MoGS1 to a 4.3 Mb interval 183 To refine the interval of interest at the MoGS1 locus, we generated subcongenic lines of the CAST 184 derived chromosomal fragments on a B6 background (N5) and tested which line suppresses ASD and 185 skeletal myopathy in Col4a1+/Δex41 mice (Fig. 3A). Line 1 which included an interval of ~10 Mb from the 186 CAST genome did not have a modifying effect on either phenotype (Fig. S1). In contrast, a subcongenic 187 line (line 2) containing the distal portion of the MoGS1 locus (from 68.7 – 73.0 Mb) showed a significant 7 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

188 protective effect for both ASD (Fig. 3B) and myopathy (Fig. 3C). While approximately 87% of eyes from 189 Col4a1+/Δex41;MoGS1B/B mice had severe ASD (1 mild, 4 moderate and 33 severe), only ~35% of the 190 eyes from Col4a1+/Δex41;MoGS1C/C mice were severe (8 mild, 14 moderate, and 12 severe). Likewise, 191 Col4a1+/Δex41;MoGS1C/C mice had significantly fewer NPN compared to Col4a1+/Δex41;MoGS1B/B mice. 192 Heterozygosity for this distal interval had intermediate effects for both phenotypes. 193 194 Candidate analysis 195 The refined MoGS1 locus is 4.3 Mb, and contains 18 protein-coding genes, and 38 predicted non- 196 protein coding genes including 13 lncRNAs, 6 snRNA genes, pseudogenes, and others (Table S1). 197 Using publicly available databases (Keane et al., 2011), we examined the refined MoGS1 interval for 198 sequence differences between the B6 and CAST genomes. Among all variations, there were 7,430 199 single nucleotide polymorphisms (SNPs), 667 small insertions and deletions (indels) and 17 large 200 structural variations including large insertions or deletions. Thirty-five SNPs were predicted to be 201 missense variants affecting 10 protein-coding genes (Table 1). Among those genes, Ankar had three 202 missense variants predicted to be deleterious by SIFT (Vaser et al., 2016) and Spag16 had one 203 predicted stop-gain variant; however, neither gene is an obvious functional candidate. Based on the 204 Database (GXD) from Mouse Genome Informatics (informatics.jax.org/expression), 205 13 of the 18 protein coding genes have been shown to be expressed both in eyes and skeletal muscles. 206 Two genes, Vwc2l and Bard1, are only expressed in eye and three genes, Ankar, Spag16, and Abca12, 207 have no reported expression in either tissue. Since both ocular and muscular defects in Col4a1+/Δex41 208 mice were suppressed by the MoGS1 locus, we hypothesized that the underlying pathogenic 209 mechanism(s) may be shared between these two tissues. Therefore, despite the presence of possible 210 pathogenic variants, we excluded Ankar, Spag16, Abca12, Vwc2I and Bard1 from further analysis. 211 212 Next, we tested whether any of the 13 genes with reported expression in ocular and muscular tissues 213 are differentially expressed in B6 mice with or without the CAST interval at the MoGS1 locus. Four 214 genes, Mreg, Fn1, Pecr1, and Igbp2, showed significantly increased expression in MoGS1C/C eyes at 215 postnatal day 0 (P0) compared to MoGS1B/B eyes by qPCR (Table 1 and Fig. 4A). FN1 is an 216 extracellular matrix protein with multiple important roles in development and tissue homeostasis by 217 interacting with cell surface receptors, extracellular matrix proteins and growth factors (Zollinger and 218 Smith, 2017). Notably, FN1 has binding sites for type IV collagen (Laurie et al., 1986; Laurie et al., 1982) 219 and localizes adjacent to basement membranes in many tissues (Laurie et al., 1982). Moreover, like 220 collagens, FN1 interacts with cells via integrin receptors (Johansson et al., 1997; Vandenberg et al., 221 1991), making it a strong functional candidate as a genetic modifier of COL4A1-related pathology. 8 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

222 223 Functional testing of FN1 and integrin signaling in MoGS1 mice 224 In silico analysis comparing the CAST and B6 alleles of Fn1 revealed 3 missense SNPs, 1 splice region 225 SNP, 7 SNPs in the UTR, 238 other noncoding SNPs, and 2 structural variations (Table S2). Murine 226 FN1 has 12 isoforms (White et al., 2008), making it difficult to predict the functional consequence(s) of 227 a particular variant. The primary consequence of Col4a1 mutations is impaired secretion of mutant 228 α1α1α2 heterotrimers into the basement membranes (Kuo et al., 2014) and it is possible that increased 229 FN1 confers partial compensation for extracellular α1α1α2 deficiency. Consistent with this hypothesis, 230 we observed increased Fn1 expression in developing eyes from MoGS1C/C mice compared to 231 MoGS1B/B mice (Fig. 4A). Furthermore, we found that FN1 levels were higher in P10 quadriceps from 232 Col4a1+/Δex41 compared to Col4a1+/+ mice and in MoGS1C/C compared to MoGS1B/B mice (Fig. 4B, C). 233 FN1 plays important roles in cell adhesion, migration and signaling during tissue morphogenesis, which 234 is primarily mediated through its interaction with integrins (Miyamoto et al., 1998; Sakai et al., 2003). 235 Therefore, we tested the effect of MoGS1 on integrin signaling by evaluating the protein levels and 236 phosphorylation status of two downstream effectors, integrin linked kinase (ILK) and focal adhesion 237 kinase (FAK) (Hu and Luo, 2013; Humphries et al., 2019) (Fig. 4B, C). Western blot analysis of P10 238 quadriceps revealed that ILK levels were higher in Col4a1+/Δex41 mice irrespective of whether they 239 carried the MoGS1B/B or MoGS1C/C interval. Similarly, FAK phosphorylation (pFAK) and pFAK/FAK ratio 240 were significantly higher in Col4a1+/Δex41;MoGS1B/B compared to Col4a1+/+;MoGS1B/B mice, and were 241 reduced by the MoGS1C/C interval. No difference was detected for the broadly used beta subunit of the 242 integrin receptor, ITGB1, between genotypes. Taken together, these data suggest that elevated integrin 243 signaling may contribute to Gould syndrome and that ASD and myopathy may be partially rescued by 244 compensatory Fn1 expression. 245 246 DISCUSSION 247 Here, we performed a genetic modifier screen in Col4a1 mutant mice to gain insight into the pathogenic 248 mechanisms that contribute to Gould syndrome. To this end, we used a large-scale F2 cross of 249 CASTxB6 and conducted independent screens for ASD and skeletal myopathy – two phenotypes that 250 are commonly observed in individuals with Gould syndrome. Independent analyses for both phenotypes 251 revealed a single, shared, semi-dominant locus on CAST Chr 1 that corroborates a previously identified 252 locus (Gould et al., 2007). To validate this genetic data, we iteratively backcrossed the locus onto the 253 B6 background and showed that a refined interval, referred to as MoGS1, suppressed ASD and 254 myopathy in Col4a1 mutant mice. However, the MoGS1 locus had no effect on porencephaly 255 penetrance and ICH severity – two phenotypes associated with Gould syndrome and previously 9 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

256 characterized in Col4a1 mutant mice. The failure of MoGS1 to suppress ICH severity was unexpected 257 since the CASTB6F1 genetic context significantly suppresses this phenotype. These data suggest that 258 the effect of the MoGS1 locus is tissue-specific and that other modifiers of ICH exist in the CAST 259 background. 260 261 Using subcongenic strains we refined the minimum critical interval for the suppressor locus to a 4.3 Mb 262 region containing 18 protein-coding genes. Of the 18 positional candidate genes, 13 are expressed 263 both in eyes and muscles, of which 6 contained missense variants between the CAST and B6 genomes; 264 however, none were predicted to be deleterious. Mutations in one of these six genes, Smarcal1, causes 265 a rare multisystem disorder (Schimke Immunoosseous dysplasia; OMIM# 242900) characterized by 266 spondyloepiphyseal dysplasia, nephrotic syndrome, T cell immunodeficiency, and a portion of patients 267 develop corneal opacity, myopia, astigmatism and optic atrophy (Boerkoel et al., 2000). SMARCAL1 268 (SWI/SNF related matrix associated, actin dependent regulator of chromatin, subfamily a-like 1) is 269 chromatin remodeling protein involved in transcriptional regulation and DNA replication, repair and 270 recombination (Bansbach et al., 2010). Compared to the B6 reference genome, CAST has two 271 missense variations in Smarcal1 that were predicted to be tolerated, and no change of expression was 272 detected in P0 eyes. Of the four positional candidate genes that were differentially expressed, Fn1 is a 273 strong functional candidate. FN1 is a ubiquitously expressed extracellular glycoprotein which plays 274 important roles in multiple processes by interacting with cell surface receptors, growth factors and other 275 extracellular matrix proteins. In cell culture, FN1 promotes collagen IV deposition, assembly and 276 incorporation into the extracellular matrix (Filla et al., 2017; Ngandu Mpoyi et al., 2020). Mice deficient 277 for Fn1 have aberrant lens placode formation and develop microphthalmia and cataracts (Hayes et al., 278 2012; Huang et al., 2011). Moreover, lack of FN1 in the skeletal muscle stem cell niche impairs muscle 279 regeneration in aged mice (Lukjanenko et al., 2016). Collectively, these findings support our genetic 280 and molecular evidence suggesting that that FN1 is a strong candidate as the gene responsible for the 281 effects conferred by the MoGS1 locus. 282 283 In general, the presence of mutant COL4A1 or COL4A2 leads to intracellular accumulation of mutant 284 heterotrimers at the expense of their secretion. Intracellular heterotrimer accumulation represents a 285 potential cell autonomous (proximal) insult (Jeanne and Gould, 2017; Mao et al., 2015). Impaired 286 heterotrimer secretion can also lead to a cell non-autonomous (distal) insult caused by extracellular 287 heterotrimer deficiency which can simultaneously perturb any number of presently unidentified cellular 288 pathways. Intracellular accumulation and extracellular deficiency can be addressed simultaneously by 289 targeting the proximal defect of protein misfolding. For example, pharmacologically promoting 10 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

290 heterotrimer secretion using a chemical chaperone, 4-phenylbutyrate (4PBA), alleviates diverse 291 pathologies in Col4a1 mutant mice (Hayashi et al., 2018; Jeanne et al., 2015; Jones et al., 2019; 292 Labelle-Dumais et al., 2019). However, even in the absence of 4PBA, secretion of mutant heterotrimers 293 is not completely abolished and the presence of mutant heterotrimers in the basement membrane 294 represents a third and distinct class of insult. We demonstrated this proof of concept by showing that 295 mice with a Col4a1G394V mutation have disproportionately severe myopathy, which is exacerbated when 296 they are treated with 4PBA (Labelle-Dumais et al., 2019). The affected residue, glycine 394, is adjacent 297 to a putative integrin-binding domain (Parkin et al., 2011) implicating impaired integrin binding in 298 skeletal myopathy. Thus, for mutations that impair subdomains of the heterotrimer responsible for 299 executing specific extracellular functions, promoting secretion of mutant heterotrimers would still be 300 predicted to ameliorate many extracellular functions. However, the increased levels of mutant 301 heterotrimers in basement membranes may also exacerbate specific disease pathway(s) related to the 302 function(s) of the impacted subdomain (Labelle-Dumais et al., 2019). Identifying the multitude of 303 extracellular roles of COL4A1/A2 and how they are executed will be key in understanding fundamental 304 aspects of matrix biology and the tissue-specific pathogenic mechanisms in Gould syndrome. Mapping 305 functional subdomains on the α1α1α2 heterotrimers will be important for genetically stratifying patients 306 and may influence the prognosis and potential therapeutic approaches. 307 308 The unbiased nature of genetic screens leaves open the possibility for finding modifiers of either 309 proximal or distal insults. By independently analyzing two phenotypes using an inbred strain (CAST) 310 with broad phenotypic suppression, we sought to find modifiers that might represent therapeutic targets 311 for the breadth of Gould Syndrome phenotypes. MoGS1, the only significant locus that we identified, 312 conferred tissue-specific effects by suppressing ASD and myopathy but not ICH. This observation is 313 consistent with a modifier that has an extracellular function and suggests that the two phenotypes have 314 at least partially overlapping distal pathogenic mechanisms. In a previous study using the Col4a1Δex41 315 mutation we found that, compared to B6, the CASTB6F1 background appeared to reduce intracellular 316 accumulation but did not increase extracellular COL4A1 levels (Jeanne et al., 2015). This observation 317 implicated intracellular accumulation in pathogenesis but did not rule out the potential of compensation 318 by other extracellular factors. The fact that MoGS1 does not suppress ICH severity indicates that 319 intracellular accumulation of mutant proteins may indeed be the primary pathogenic insult for that 320 phenotype. 321 322 When we compared disease severity across a murine allelic series of 9 Col4a1 and Col4a2 mutations, 323 we found that the Col4a1G394V mutation had relatively little intracellular accumulation and that 11 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

324 Col4a1+/G394V mice had mild ICH but severe ASD and myopathy (Kuo et al., 2014). The proximity of the 325 Col4a1G394V mutation to a predicted integrin binding domain suggests that impaired integrin binding 326 contributes to ocular dysgenesis and myopathy. Therefore, one possibility is that FN1 ameliorates ASD 327 and skeletal myopathy caused by Col4a1 mutations through compensatory integrin-mediated 328 interactions. Consistent with this hypothesis, we detected evidence for increased integrin signaling in 329 Col4a1+/Δex41 mice that was reduced by the MoGS1 locus from CAST in the context of increased Fn1 330 expression. Moreover, ILK and FAK are both previously implicated in lens development (Cammas et 331 al., 2012; Harburger and Calderwood, 2009; Hayes et al., 2012; Samuelsson et al., 2007; Teo et al., 332 2014) and myopathy (Boppart and Mahmassani, 2019; Gheyara et al., 2007). Functional validation of 333 this pathway with experimental manipulation and preclinical interventions would represent a significant 334 advance in the understanding of the pathogenic mechanisms that contribute to Gould syndrome. 335 336 MATERIALS AND METHODS 337 Mice 338 All experiments were compliant with the ARVO Statement for the Use of Animals in Ophthalmic and 339 Vision Research and approved by the Institutional Animal Care and Use Committee at the University 340 of California, San Francisco. Col4a1+/Δex41 mice were originally identified in a mutagenesis screen 341 conducted at The Jackson Laboratory (Bar Harbor, ME) (Gould et al., 2007; Gould et al., 2005). Since 342 then Col4a1+/Δex41 mice has been backcrossed to B6 for at least 20 generations. CAST mice were 343 obtained from The Jackson Laboratory. All animals were maintained in full-barrier facilities free of 344 specific pathogens on a 12-hour light/dark cycle with food and water ad libitum. Both male and female 345 mice were used in this study. 346 347 Slit lamp biomicroscopy 348 Ocular anterior segment examinations were performed on mice at 1.0- 1.5 months of age using a slit 349 lamp biomicroscope (Topcon SL-D7; Topcon Medical Systems, Oakland, NJ) attached to a digital SLR 350 camera (Nikon D200; Nikon, Melville, NY). 351 352 Muscle analysis 353 At 2 months of age, mice were subjected to treadmill exercise and their quadriceps were harvested two 354 days later. Exercise was 30 minutes with a 15° downhill grade on a treadmill equipped with a shock 355 plate (Columbus Instruments, Columbus, OH). Animals were started at 6 m/min and increased by 3 356 m/min every 2 min until maximum of 15 m/min speed was reached. Quadriceps were dissected and 357 frozen in liquid nitrogen cooled isopentane. Cryosections (10 μm) were collected at regular intervals 12 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

358 and stained with hematoxylin and eosin (H&E) for histopathology and the number of NPN were 359 evaluated on a total of 12 sections for each muscle and the percentage of muscle fibers with NPN was 360 quantified. The observers were masked to genotypes and counted between 2000–5000 muscle fibers 361 per animal. 362 363 Brain histological analysis 364 ICH was assessed at 2 months of age by Perl’s Prussian blue staining as previously described (Jeanne 365 et al., 2015). Briefly, mice underwent transcardial perfusion with saline followed by 4% 366 paraformaldehyde (PFA) and then posted fixed in 4% PFA overnight. Brains were dissected, 367 cryoprotected in 30% sucrose and embedded in optimal cutting temperature compound (Sakura 368 Finetek, Torrance, CA). Coronal cryo-sections (35 μm) regularly spaced along the rostro-caudal axis 369 were stained with Prussian blue/Fast red and imaged. On each section, the percentage of brain area 370 with Prussian blue staining was calculated using ImageJ software (National Institutes of Health). ICH 371 severity was expressed as the average percentage of hemosiderin surface area on 28 sections for 372 each brain. The presence or absence of porencephaly on sections used in ICH analysis was also 373 recorded. 374 375 Genome scan +/Δex41 376 Genomic DNA was obtained from 192 (CAST X B6) F2 mice carrying the Col4a1 mutation and 377 genotyped at the UCSF Genomic Core Facility with a commercial single nucleotide polymorphism (SNP)

378 panel (Illumina, San Diego, CA) which contains 646 informative SNPs for (CAST X B6) F2 progeny. A 379 genome-wide suppressor screen was performed in R (R Core Team, 2019) using the package R/qtl 380 (Broman et al., 2003) treating the trait as binary. Genome-wide significance was established using 1000 381 permutation testing (Churchill and Doerge, 1994). A confidence interval for each QTL location was then 382 calculated using Bayesian confidence sets (Sen and Churchill, 2001). The ASD score for each animal 383 was averaged from both eyes and treated as a dichotomous variable where score of 1 was mild, and 384 score of 1.5 -3 was severe. A subset (49) of these 192 mice was randomly selected in a separate 385 mapping analysis for the muscle modifier as a quantitative trait. 386

387 Quantitative polymerase chain reaction (qPCR)

388 Eyes from P0 mice were dissected in PBS, immediately transferred in RNAlater (ThermoFisher 389 Scientific, USA) and stored at -80°C until use. Total RNA was extracted using TRIzol Reagent 390 (ThermoFisher Scientific) according to the manufacturer’s instructions and reverse transcribed to cDNA

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

391 using iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). Quantitative PCR was performed on a Bio- 392 Rad CFX96 real-time system using SsoFast Evagreen mix (Bio-Rad) and primers are listed in Table 393 S3. Briefly, 10 ng of cDNA and 1.25 μM primers were used per reaction in a final volume of 10 μl. Each 394 cycle consisted of denaturation at 95°C for 5s, followed by annealing and extension at 60°C for 5s for 395 a total of 45 cycles. All experiments were run with technical duplicates and 4-6 biological replicates 396 were used per group. The relative expression of each gene was normalized to Hprt1 or Gapdh and 397 analyzed using the 2-∆∆CT method (Livak and Schmittgen, 2001). 398 399 Western blot analyses 400 Quadriceps from P10 pups were dissected and total proteins extracted using radioimmunoprecipitation 401 assay (RIPA) buffer (VWR, Radnor, PA) supplemented with Halt Protease and Phosphatase Inhibitor 402 Cocktail (ThermoFisher Scientific), EDTA and 2 mM phenylmethylsulfonyl fluoride. Total proteins (10 403 μg) were separated on 4–15% gradient SDS-PAGE gels (Bio-Rad) and transferred to polyvinylidene 404 fluoride (PVDF) membranes (BioRad). Membranes were blocked for 1 hr at room temperature in 5% 405 non-fat milk in TBS containing 0.1% Tween-20 (TBST), incubated overnight at 4°C in primary antibodies 406 diluted in 2% non-fat milk in TBST. Primary antibodies and dilutions used are as follow: rabbit anti-FN1 407 (Abcam #Ab2413 1:10000, Abcam), rabbit anti-pFAK (1:1000, Invitrogen #44624G), rabbit anti-FAK 408 (1:1000, Santa Cruz #SC-557), rabbit anti-ILK (1:1000, Cell Signaling Technology #3862), mouse anti- 409 ITGB1 (1:1000, BD Biosciences #610467), and mouse anti-GAPDH (1:500000, Millipore #MAB374). 410 Following washes in TBST, membranes were incubated for 1 hr at room temperature with species- 411 specific horseradish peroxidase conjugated secondary antibodies (1:10000, Jackson ImmunoResearch, 412 West Grove, PA) diluted in 2% non-fat milk in TBST. Immunoreactivity was visualized using 413 chemiluminescence (ECL or Luminata Forte substrate, ThermoFisher Scientific) and imaged using a 414 ChemiDoc MP Imaging System (Bio-Rad) or exposed to X-ray films. Densitometric analyses were 415 performed on low exposure images using the Quantity One software (Bio-Rad). 416 417 Statistics 418 Statistical analyses and graphs were prepared using GraphPad Prism v8.0 Software (GraphPad, La 419 Jolla, CA). Multiple comparisons were carried using one-way ANOVA followed by Tukey post-test or 420 Kruskal-Wallis test followed by Dunn’s post-test. p values less than 0.05 were considered statistically 421 significant. 422 423 ACKNOWLEDGEMENTS

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

424 We thank members of the Gould Laboratory for their critical reviews of the manuscript. We also thank 425 Cassandre Labelle-Dumais for helpful discussion and guidance with the muscle studies. 426 427 COMPETING INTERESTS 428 M.J. is an employee of Genentech, Inc., a member of the Roche group. K.H. is an employee of 429 Centrillion Technologies. 430 431 FUNDING 432 This work was supported by NIH grants R01EY019887 (DBG), R01NS096173 (DBG), Research to 433 Prevent Blindness (DBG), That Man May See (MM and DBG), and Knights Templar Eye Foundation 434 (MM) and Bright Focus (DBG). This research was supported, in part, by the UCSF Vision Core shared 435 resource of the NIH/NEI P30 EY002162, and by an unrestricted grant from Research to Prevent 436 Blindness, New York, NY. 437 438 AUTHOR CONTRIBUTIONS 439 Conceived and designed the experiments: D.B.G and M.M. Performed experiments: M.M, T.P, M.J, 440 and K.H. Data analysis: M.M, T.P, M.J, K.H, S.S and D.B.G. Contributed reagents, materials, and/or 441 analysis tools: S.S. Writing of the manuscript: M.M and D.B.G.

15 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

442 Table 1. List of protein-coding genes in the refined MoGS locus 443

Entrez # of # of # of stop- # of SNPs in # of Relative Gene Eye Muscle gene missense deleterious gain splice SNPs in mRNA p - value name expression expression ID SNPs SNPs SNPs region 3'UTR expression

Erbb4 13869 0 0 0 0 0 Yes (> E18) Yes (< E18) 0.88 0.559

Ikzf2 22779 2 0 0 0 40 Yes (> P4) Yes (> P4) 0.79 0.054

Spag16 66722 8 0 1 5 39 - - NA NA

Vwc2l 320460 0 0 0 0 21 Yes (> P4) - NA NA

Bard1 12021 3 0 0 0 29 Yes (> P4) - NA NA

Abca12 74591 5 0 0 4 4 - - NA NA

Atic 108147 3 0 0 2 17 Yes (> E15) Yes (> P4) 0.82 0.267

Fn1 14268 3 0 0 10 5 Yes (> E8) Yes (> E13) 1.45 0.007**

Mreg 381269 0 0 0 0 16 Yes (> P4) Yes (> P4) 1.43 0.026*

Pecr 111175 0 0 0 1 8 Yes (> P4) Yes (> P4) 0.80 0.026*

Tmem169 271711 1 0 0 1 7 Yes (> P4) Yes (> P4) 0.97 0.724

Xrcc5 22596 3 0 0 5 5 Yes (> P4) Yes (> P4) 0.88 0.496

March4 381270 0 0 0 0 13 Yes (> P4) Yes (> P4) 1.20 0.452

Smarcal1 54380 2 0 0 0 0 Yes (> P4) Yes (> P4) 0.95 0.459

Ankar 319695 5 3 0 0 0 - - NA NA

Rpl37a 19981 0 0 0 2 1 Yes (> P4) Yes (> P4) 1.07 0.192

Igfbp2 16008 0 0 0 0 1 Yes (> E13) Yes (> P4) 0.83 0.023*

Igfbp5 16011 0 0 0 0 36 Yes (> E13) Yes (> E15) 1.34 0.076 444 E, embryonic day; P, postnatal; -, expression not detectable; NA, data not available; * p < 0.05; ** p < 0.01 by Student's t-test 445 446

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

447 REFERENCES 448 Bansbach, C. E., Boerkoel, C. F. and Cortez, D. (2010). SMARCAL1 and replication stress: 449 an explanation for SIOD? Nucleus 1, 245-8. 450 Bilguvar, K., DiLuna, M. L., Bizzarro, M. J., Bayri, Y., Schneider, K. C., Lifton, R. P., 451 Gunel, M. and Ment, L. R. (2009). COL4A1 mutation in preterm intraventricular hemorrhage. J 452 Pediatr 155, 743-5. 453 Boerkoel, C. F., O'Neill, S., André, J. L., Benke, P. J., Bogdanovíć, R., Bulla, M., Burguet, 454 A., Cockfield, S., Cordeiro, I., Ehrich, J. H. et al. (2000). Manifestations and treatment of Schimke 455 immuno-osseous dysplasia: 14 new cases and a review of the literature. Eur J Pediatr 159, 1-7. 456 Boppart, M. D. and Mahmassani, Z. S. (2019). Integrin signaling: linking mechanical 457 stimulation to skeletal muscle hypertrophy. Am J Physiol Cell Physiol 317, C629-C641. 458 Breedveld, G., de Coo, I. F., Lequin, M. H., Arts, W. F., Heutink, P., Gould, D. B., John, S. 459 W., Oostra, B. and Mancini, G. M. (2006). Novel mutations in three families confirm a major role of 460 COL4A1 in hereditary porencephaly. J Med Genet 43, 490-5. 461 Broman, K. W., Wu, H., Sen, S. and Churchill, G. A. (2003). R/qtl: QTL mapping in 462 experimental crosses. Bioinformatics 19, 889-90. 463 Cammas, L., Wolfe, J., Choi, S. Y., Dedhar, S. and Beggs, H. E. (2012). Integrin-linked 464 kinase deletion in the developing lens leads to capsule rupture, impaired fiber migration and non- 465 apoptotic epithelial cell death. Invest Ophthalmol Vis Sci 53, 3067-81. 466 Ceco, E. and McNally, E. M. (2013). Modifying muscular dystrophy through transforming 467 growth factor-β. FEBS J 280, 4198-209. 468 Churchill, G. A. and Doerge, R. W. (1994). Empirical threshold values for quantitative trait 469 mapping. Genetics 138, 963-71. 470 Coupry, I., Sibon, I., Mortemousque, B., Rouanet, F., Mine, M. and Goizet, C. (2010). 471 Ophthalmological features associated with COL4A1 mutations. Arch Ophthalmol 128, 483-9. 472 de Vries, L. S., Koopman, C., Groenendaal, F., Van Schooneveld, M., Verheijen, F. W., 473 Verbeek, E., Witkamp, T. D., van der Worp, H. B. and Mancini, G. (2009). COL4A1 mutation in two 474 preterm siblings with antenatal onset of parenchymal hemorrhage. Ann Neurol 65, 12-8. 475 Favor, J., Gloeckner, C. J., Janik, D., Klempt, M., Neuhauser-Klaus, A., Pretsch, W., 476 Schmahl, W. and Quintanilla-Fend, L. (2007). Type IV procollagen missense mutations associated 477 with defects of the eye, vascular stability, the brain, kidney function and embryonic or postnatal 478 viability in the mouse, Mus musculus: an extension of the Col4a1 allelic series and the identification of 479 the first two Col4a2 mutant alleles. Genetics 175, 725-36. 480 Fidler, A. L., Darris, C. E., Chetyrkin, S. V., Pedchenko, V. K., Boudko, S. P., Brown, K. L., 481 Gray Jerome, W., Hudson, J. K., Rokas, A. and Hudson, B. G. (2017). Collagen IV and basement 482 membrane at the evolutionary dawn of metazoan tissues. Elife 6. 483 Filla, M. S., Dimeo, K. D., Tong, T. and Peters, D. M. (2017). Disruption of fibronectin matrix 484 affects type IV collagen, fibrillin and laminin deposition into extracellular matrix of human trabecular 485 meshwork (HTM) cells. Exp Eye Res 165, 7-19. 486 Gheyara, A. L., Vallejo-Illarramendi, A., Zang, K., Mei, L., St-Arnaud, R., Dedhar, S. and 487 Reichardt, L. F. (2007). Deletion of integrin-linked kinase from skeletal muscles of mice resembles 488 muscular dystrophy due to alpha 7 beta 1-integrin deficiency. Am J Pathol 171, 1966-77. 489 Giorgio, E., Vaula, G., Bosco, G., Giacone, S., Mancini, C., Calcia, A., Cavalieri, S., Di 490 Gregorio, E., Rigault De Longrais, R., Leombruni, S. et al. (2015). Two families with novel 491 missense mutations in COL4A1: When diagnosis can be missed. J Neurol Sci 352, 99-104. 492 Gould, D. B., Marchant, J. K., Savinova, O. V., Smith, R. S. and John, S. W. (2007). Col4a1 493 mutation causes endoplasmic reticulum stress and genetically modifiable ocular dysgenesis. Hum 494 Mol Genet 16, 798-807.

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

495 Gould, D. B., Phalan, F. C., Breedveld, G. J., van Mil, S. E., Smith, R. S., Schimenti, J. C., 496 Aguglia, U., van der Knaap, M. S., Heutink, P. and John, S. W. (2005). Mutations in Col4a1 cause 497 perinatal cerebral hemorrhage and porencephaly. Science 308, 1167-71. 498 Gould, D. B., Phalan, F. C., van Mil, S. E., Sundberg, J. P., Vahedi, K., Massin, P., 499 Bousser, M. G., Heutink, P., Miner, J. H., Tournier-Lasserve, E. et al. (2006). Role of COL4A1 in 500 small-vessel disease and hemorrhagic stroke. N Engl J Med 354, 1489-96. 501 Harburger, D. S. and Calderwood, D. A. (2009). Integrin signalling at a glance. J Cell Sci 502 122, 159-63. 503 Hayashi, G., Labelle-Dumais, C. and Gould, D. B. (2018). Use of sodium 4-phenylbutyrate to 504 define therapeutic parameters for reducing intracerebral hemorrhage and myopathy in Col4a1 mutant 505 mice. Dis Model Mech 11. 506 Hayes, J. M., Hartsock, A., Clark, B. S., Napier, H. R., Link, B. A. and Gross, J. M. (2012). 507 Integrin alpha5/fibronectin1 and focal adhesion kinase are required for lens fiber morphogenesis in 508 zebrafish. Mol Biol Cell 23, 4725-38. 509 Hu, P. and Luo, B. H. (2013). Integrin bi-directional signaling across the plasma membrane. J 510 Cell Physiol 228, 306-12. 511 Huang, J., Rajagopal, R., Liu, Y., Dattilo, L. K., Shaham, O., Ashery-Padan, R. and Beebe, 512 D. C. (2011). The mechanism of lens placode formation: a case of matrix-mediated morphogenesis. 513 Dev Biol 355, 32-42. 514 Humphries, J. D., Chastney, M. R., Askari, J. A. and Humphries, M. J. (2019). Signal 515 transduction via integrin adhesion complexes. Curr Opin Cell Biol 56, 14-21. 516 Jeanne, M. and Gould, D. B. (2017). Genotype-phenotype correlations in pathology caused 517 by collagen type IV alpha 1 and 2 mutations. Matrix Biol 57-58, 29-44. 518 Jeanne, M., Jorgensen, J. and Gould, D. B. (2015). Molecular and Genetic Analyses of 519 Collagen Type IV Mutant Mouse Models of Spontaneous Intracerebral Hemorrhage Identify 520 Mechanisms for Stroke Prevention. Circulation 131, 1555-65. 521 Johansson, S., Svineng, G., Wennerberg, K., Armulik, A. and Lohikangas, L. (1997). 522 Fibronectin-integrin interactions. Front Biosci 2, d126-46. 523 Jones, F. E., Murray, L. S., McNeilly, S., Dean, A., Aman, A., Lu, Y., Nikolova, N., 524 Malomgre, R., Horsburgh, K., Holmes, W. M. et al. (2019). 4-Sodium phenyl butyric acid has both 525 efficacy and counter-indicative effects in the treatment of Col4a1 disease. Hum Mol Genet 28, 628- 526 638. 527 Keane, T. M., Goodstadt, L., Danecek, P., White, M. A., Wong, K., Yalcin, B., Heger, A., 528 Agam, A., Slater, G., Goodson, M. et al. (2011). Mouse genomic variation and its effect on 529 phenotypes and gene regulation. Nature 477, 289-94. 530 Kuo, D. S., Labelle-Dumais, C., Mao, M., Jeanne, M., Kauffman, W. B., Allen, J., Favor, J. 531 and Gould, D. B. (2014). Allelic heterogeneity contributes to variability in ocular dysgenesis, 532 myopathy and brain malformations caused by Col4a1 and Col4a2 mutations. Hum Mol Genet 23, 533 1709-22. 534 Labelle-Dumais, C., Dilworth, D. J., Harrington, E. P., de Leau, M., Lyons, D., Kabaeva, 535 Z., Manzini, M. C., Dobyns, W. B., Walsh, C. A., Michele, D. E. et al. (2011). COL4A1 mutations 536 cause ocular dysgenesis, neuronal localization defects, and myopathy in mice and Walker-Warburg 537 syndrome in humans. PLoS Genet 7, e1002062. 538 Labelle-Dumais, C., Schuitema, V., Hayashi, G., Hoff, K., Gong, W., Dao, D. Q., Ullian, E. 539 M., Oishi, P., Margeta, M. and Gould, D. B. (2019). COL4A1 Mutations Cause Neuromuscular 540 Disease with Tissue-Specific Mechanistic Heterogeneity. Am J Hum Genet 104, 847-860. 541 Laurie, G. W., Bing, J. T., Kleinman, H. K., Hassell, J. R., Aumailley, M., Martin, G. R. and 542 Feldmann, R. J. (1986). Localization of binding sites for laminin, heparan sulfate proteoglycan and 543 fibronectin on basement membrane (type IV) collagen. J Mol Biol 189, 205-16.

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

544 Laurie, G. W., Leblond, C. P. and Martin, G. R. (1982). Localization of type IV collagen, 545 laminin, heparan sulfate proteoglycan, and fibronectin to the basal lamina of basement membranes. J 546 Cell Biol 95, 340-4. 547 Livak, K. J. and Schmittgen, T. D. (2001). Analysis of relative gene expression data using 548 real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25, 402-8. 549 Lukjanenko, L., Jung, M. J., Hegde, N., Perruisseau-Carrier, C., Migliavacca, E., Rozo, M., 550 Karaz, S., Jacot, G., Schmidt, M., Li, L. et al. (2016). Loss of fibronectin from the aged stem cell 551 niche affects the regenerative capacity of skeletal muscle in mice. Nat Med 22, 897-905. 552 Mao, M., Alavi, M. V., Labelle-Dumais, C. and Gould, D. B. (2015). Type IV Collagens and 553 Basement Membrane Diseases: Cell Biology and Pathogenic Mechanisms. Curr Top Membr 76, 61- 554 116. 555 Meuwissen, M. E., Halley, D. J., Smit, L. S., Lequin, M. H., Cobben, J. M., de Coo, R., van 556 Harssel, J., Sallevelt, S., Woldringh, G., van der Knaap, M. S. et al. (2015). The expanding 557 phenotype of COL4A1 and COL4A2 mutations: clinical data on 13 newly identified families and a 558 review of the literature. Genet Med 17, 843-53. 559 Meyer, K. J. and Anderson, M. G. (2017). Genetic modifiers as relevant biological variables 560 of eye disorders. Hum Mol Genet 26, R58-r67. 561 Miyamoto, S., Katz, B. Z., Lafrenie, R. M. and Yamada, K. M. (1998). Fibronectin and 562 integrins in cell adhesion, signaling, and morphogenesis. Ann N Y Acad Sci 857, 119-29. 563 Ngandu Mpoyi, E., Cantini, M., Sin, Y. Y., Fleming, L., Zhou, D. W., Costell, M., Lu, Y., 564 Kadler, K., Garcia, A. J., Van Agtmael, T. et al. (2020). Material-driven fibronectin assembly 565 rescues matrix defects due to mutations in collagen IV in fibroblasts. Biomaterials 252, 120090. 566 Parkin, J. D., San Antonio, J. D., Pedchenko, V., Hudson, B., Jensen, S. T. and Savige, J. 567 (2011). Mapping structural landmarks, ligand binding sites, and missense mutations to the collagen IV 568 heterotrimers predicts major functional domains, novel interactions, and variation in phenotypes in 569 inherited diseases affecting basement membranes. Hum Mutat 32, 127-43. 570 Poschl, E., Schlotzer-Schrehardt, U., Brachvogel, B., Saito, K., Ninomiya, Y. and Mayer, 571 U. (2004). Collagen IV is essential for basement membrane stability but dispensable for initiation of its 572 assembly during early development. Development 131, 1619-28. 573 R Core Team. (2019). R: A language and environment for statistical 574 computing. Vienna, Austria: R Foundation for Statistical Computing. 575 Ricard-Blum, S. (2011). The collagen family. Cold Spring Harb Perspect Biol 3, a004978. 576 Rødahl, E., Knappskog, P. M., Majewski, J., Johansson, S., Telstad, W., Kråkenes, J. and 577 Boman, H. (2013). Variants of anterior segment dysgenesis and cerebral involvement in a large 578 family with a novel COL4A1 mutation. Am J Ophthalmol 155, 946-53. 579 Sakai, T., Li, S., Docheva, D., Grashoff, C., Sakai, K., Kostka, G., Braun, A., Pfeifer, A., 580 Yurchenco, P. D. and Fassler, R. (2003). Integrin-linked kinase (ILK) is required for polarizing the 581 epiblast, cell adhesion, and controlling actin accumulation. Genes Dev 17, 926-40. 582 Samuelsson, A. R., Belvindrah, R., Wu, C., Muller, U. and Halfter, W. (2007). Beta1-integrin 583 signaling is essential for lens fiber survival. Gene Regul Syst Bio 1, 177-89. 584 Sen, S. and Churchill, G. A. (2001). A statistical framework for quantitative trait mapping. 585 Genetics 159, 371-87. 586 Shah, S., Ellard, S., Kneen, R., Lim, M., Osborne, N., Rankin, J., Stoodley, N., van der 587 Knaap, M., Whitney, A. and Jardine, P. (2012). Childhood presentation of COL4A1 mutations. Dev 588 Med Child Neurol 54, 569-74. 589 Sibon, I., Coupry, I., Menegon, P., Bouchet, J. P., Gorry, P., Burgelin, I., Calvas, P., 590 Orignac, I., Dousset, V., Lacombe, D. et al. (2007). COL4A1 mutation in Axenfeld-Rieger anomaly 591 with leukoencephalopathy and stroke. Ann Neurol 62, 177-84. 592 Teo, Z. L., McQueen-Miscamble, L., Turner, K., Martinez, G., Madakashira, B., Dedhar, S., 593 Robinson, M. L. and de Iongh, R. U. (2014). Integrin linked kinase (ILK) is required for lens 594 epithelial cell survival, proliferation and differentiation. Exp Eye Res 121, 130-42. 19 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

595 Thaung, C., West, K., Clark, B. J., McKie, L., Morgan, J. E., Arnold, K., Nolan, P. M., 596 Peters, J., Hunter, A. J., Brown, S. D. et al. (2002). Novel ENU-induced eye mutations in the 597 mouse: models for human eye disease. Hum Mol Genet 11, 755-67. 598 Vahedi, K., Kubis, N., Boukobza, M., Arnoult, M., Massin, P., Tournier-Lasserve, E. and 599 Bousser, M. G. (2007). COL4A1 mutation in a patient with sporadic, recurrent intracerebral 600 hemorrhage. Stroke 38, 1461-4. 601 Vandenberg, P., Kern, A., Ries, A., Luckenbill-Edds, L., Mann, K. and Kuhn, K. (1991). 602 Characterization of a type IV collagen major cell binding site with affinity to the alpha 1 beta 1 and the 603 alpha 2 beta 1 integrins. J Cell Biol 113, 1475-83. 604 Vaser, R., Adusumalli, S., Leng, S. N., Sikic, M. and Ng, P. C. (2016). SIFT missense 605 predictions for genomes. Nat Protoc 11, 1-9. 606 Vieira, N. M., Elvers, I., Alexander, M. S., Moreira, Y. B., Eran, A., Gomes, J. P., Marshall, 607 J. L., Karlsson, E. K., Verjovski-Almeida, S., Lindblad-Toh, K. et al. (2015). Jagged 1 Rescues 608 the Duchenne Muscular Dystrophy Phenotype. Cell 163, 1204-1213. 609 Vilain, C., Van Regemorter, N., Verloes, A., David, P. and Van Bogaert, P. (2002). 610 Neuroimaging fails to identify asymptomatic carriers of familial porencephaly. Am J Med Genet 112, 611 198-202. 612 White, E. S., Baralle, F. E. and Muro, A. F. (2008). New insights into form and function of 613 fibronectin splice variants. J Pathol 216, 1-14. 614 Yoneda, Y., Haginoya, K., Kato, M., Osaka, H., Yokochi, K., Arai, H., Kakita, A., 615 Yamamoto, T., Otsuki, Y., Shimizu, S. et al. (2013). Phenotypic spectrum of COL4A1 mutations: 616 porencephaly to schizencephaly. Ann Neurol 73, 48-57. 617 Zagaglia, S., Selch, C., Nisevic, J. R., Mei, D., Michalak, Z., Hernandez-Hernandez, L., 618 Krithika, S., Vezyroglou, K., Varadkar, S. M., Pepler, A. et al. (2018). Neurologic phenotypes 619 associated with COL4A1/2 mutations: Expanding the spectrum of disease. Neurology 91, e2078- 620 e2088. 621 Zollinger, A. J. and Smith, M. L. (2017). Fibronectin, the extracellular glue. Matrix Biol 60-61, 622 27-37. 623 624

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

625 FIGURE LEGENDS 626 Figure 1. A genome-wide screen identified a single modifier locus on Chr 1 for ASD and

627 myopathy. (A) Schematic representation of the (CAST X B6) F2 cross. One hundred ninety two mutant +/Δex41 628 (CAST X B6-Col4a1 ) F2 progeny were genotyped and phenotyped for ASD. (B) Representative 629 slit-lamp images illustrating scoring of ASD severity: mild, score = 1; moderate, score = 2; severe, score 630 = 3. (C) A one-dimensional genome scan identified a locus on Chr 1 for ASD (max. LOD score = 11.2, 631 99% Bayesian confidence interval extending from 51.3 - 73.0 Mb, Ensembl GRCm38.p6). Solid 632 horizontal line, 5% false positive threshold; dashed horizontal line, 10% threshold. (D) A genome-wide 633 scan conditioned on the Chr 1 modifier locus suggests that no other single locus has a strong effect on 634 ASD. Dashed horizontal line: 20% false positive threshold. (E) Representative images of cross sections 635 from quadriceps stained with H&E showing variable myopathy severity in Col4a1+/Δex41 mice with 636 different genetic backgrounds. Arrows indicate muscle fibers with NPN. Scale bar = 20 μm. (F) To

637 identify skeletal muscle modifier loci, a subset (49) of the 192 (CAST X B6) F2 progeny was assessed 638 for myopathy. One dimensional genome scan identified a myopathy modifier at the same position as 639 the ASD modifier (Max. LOD score = 5.61). 640 641 Figure 2. An incipient congenic strain containing the CAST-derived MoGS1 locus suppressed 642 ASD and myopathy, but not ICH. (A) Evaluation of eyes from the incipient congenic mice at N5F2 643 validates MoGS1 as a genetic suppressor of ASD. The bar graph shows the percentage of eyes at 644 each ASD severity level from Col4a1+/Δex41 mice that are homozygous for B6 alleles (B/B), heterozygous 645 or homozygous for CAST alleles (C/B and C/C, respectively) in the congenic interval. ASD for 646 Col4a1+/Δex41 mice on the pure B6 and CASTB6F1 backgrounds are also shown. *p < 0.05; ****p < 647 0.0001, by Kruskal-Wallis test followed by Dunn’s multiple comparison test. (B) Evaluation of muscles 648 from the N5F2 congenic mice validates MoGS1 as a genetic suppressor of skeletal myopathy. Data 649 are presented as mean ± SEM. *p < 0.05, by one-way ANOVA followed by Tukey’s multiple comparison 650 test. (C) Penetrance of porencephaly for Col4a1+/Δex41 mice with different genetic backgrounds was 651 determined by brain histology. Top representative images show absence or presence of porencephaly. 652 * marks brain area with porencephaly. (D) ICH in mice was assessed using Perl’s Prussian blue staining. 653 Data for each mouse are shown as a percentage of brain area with Prussian blue staining averaged 654 over 28 sections from each brain. Representative images of Prussian blue-stained brain sections and 655 respective Prussian blue staining quantification are shown above the graphs. Sample sizes are 656 indicated in parentheses. Data are presented as mean ± SEM. *p < 0.05, Student’s t-test for comparison 657 between B6 and CASTB6F1; One-way ANOVA followed by Tukey’s test for multiple comparisons 658 between B/B, C/B, and C/C. 21 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

659 660 Figure 3. Fine mapping of the MoGS1 locus. (A) Schematic illustration of subcongenic lines 661 containing different intervals of the CAST modifier locus backcrossed onto the B6 background for 5 662 generations. Subcongenic Line 2 (68723089 – 72973981bp on Chr 1, Ensembl GRCm38.p6) 663 ameliorated ASD and myopathy severity. Black line, chromosome fragments from B6; grey bar, 664 chromosome fragments from CAST. CI, critical interval. (B) Analyses of ASD severity in mice 665 homozygous for B6 alleles (B/B), heterozygous or homozygous for CAST alleles (C/B and C/C, 666 respectively) at the refined distal interval. Percentage of eyes at each ASD severity level was shown. 667 ***p<0.001 by Kruskal-Wallis test followed by Dunn’s multiple comparison test. (C) Analysis of 668 myopathy by NPN showing severity in mice with different genotypes at the refined distal interval. 669 Sample size indicated in parentheses. Data are presented as mean ± SEM.*p < 0.05, one-way ANOVA 670 followed by Tukey’s multiple comparison test. 671 672 Figure 4. Evaluating FN1 as a candidate modifier gene. (A) Quantitative qPCR analysis showed 673 increased expression of Fn1 mRNA in P0 eyes from Col4a1+/+ mice homozygous for the CAST interval 674 at the refined MoGS1 locus (+/+;C/C) compared to mice homozygous for the B6 interval (+/+;B/B). n = 675 4 per genotype. Data are presented as mean ± SEM. *p < 0.05, by Student’s t-test. (B-C) 676 Representative images (B) and quantification (C) of Western-blot analyses of P10 quadriceps showing 677 a trend towards increased FN1 protein levels in mice homozygous for CAST alleles at the refined 678 MoGS1 locus (+/+;C/C and +/Δex41;C/C) compared to mice homozygous for B6 alleles (+/+;B/B and 679 +/Δex41;B/B). ILK levels were increased in Col4a1+/Δex41 mice irrespective of the genotype at the 680 MoGS1 locus (+/Δex41;B/B and +/Δex41;C/C, p = 0.06 for MoGS1C/C) compared to their corresponding 681 Col4a1+/+ controls (+/+;B/B, and +/+;C/C, respectively). Ratios of pFAK/FAK, were significantly elevated 682 in Col4a1+/Δex41 mice homozygous for MoGS1 B6 alleles (+/Δex41;B/B) but not CAST alleles 683 (+/Δex41;C/C) (p = 0.35) compared to their corresponding Col4a1+/+ controls (+/+;B/B, and +/+;C/C, 684 respectively) suggesting altered integrin signaling that is partially restored by the MoGS1 locus. Levels 685 of ITBG1 did not differ between genotypes. n = 4, 5, 5, 5 for +/+;B/B, +/Δex41;B/B, +/+;C/C and 686 +/Δex41;C/C respectively. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01 by one-way 687 ANOVA followed by Sidak’s multiple comparison test. 688 689 Figure S1. Subcongenic Line 1 did not confer the modifier effect. (A-B) A ~10 Mb proximal portion 690 of the CAST ModGS1 locus (52418578-61808145bp on Chr 1, Ensembl GRCm38.p6) showed no 691 protective effect for ASD (A) or skeletal myopathy (B) phenotypes seen in Col4a1+/Δex41 mice. Sample

22 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

692 size indicated in parentheses. Data are presented as mean ± SEM. one-way ANOVA followed by 693 Tukey’s multiple comparison test. 694 695 Table S1. List of protein-coding and non-coding genes in the ModGS1 locus. 696 697 Table S2. SNPs, small InDels and structural variants in Fn1. 698 699 Table S3. Primer sequences for quantitative PCR.

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

700

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 1 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.06.371666; this version posted November 7, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 4